Guangdong province in China is the latest place to be hit by formerly-unusual torrential rains.

Will this teach China that continuing to increase its greenhouse emissions would be deadly?

Salafi Arabia is running a marketing campaign to pretend it is something other than repressive and sexist.

Due to global heating, *northern permafrost region [now] emits more greenhouse gases than it captures.*

This is an example of a positive feedback. Positive feedbacks, which we can't entirely predict, are what can destroy civilization.

Some hospitals in abortion-forbidding states turn women away from emergency rooms so as not to get stuck choosing between the crime of saving them or the liability of letting them die.

The US government s trying to enforce the law that emergency rooms can't just send a patient away. Now the case has come to the Supreme Court, where right-wing fanatics may decide to kill women in the name of the sacredness of fetuses.

The new president of the World Bank wants funds to invest in decrbonization of poor countries.

This lending program would be a good idea, but the highest priorities for investments in decarbonization are:

1. Getting wealthier countries to invest in their own decarbonization, They produce a lot more greenhouse gas than poor counries do. They do invest in renewable electric generation, but their fossil fuel extraction is still increasing year by year, and greenhouse emissions likewise.
2. Putting an end to funding for "aid" that takes the form of "investment" and "development" of fossil fuel extraction, transport and combustion in poor countries.

Shifting those funds to decarbonization aid would be the best way to start that.

But if we don't require these actions. decarbonization won't get off the ground.

Yanis Varoufakis presents the text of the speech over which Germany banned his presence and even from presenting videos of his views. This puts political freedom in Germany under grave threat.

In the text he affirms his solidarity with all victims of atrocities, including Jews and Palestinians. I don't agree with every detail of his views, but rather than quibble, I say his heart is in the right place.

Zoom demands that users make accounts and identify themselves before they can participate in a call. This puts Zoom in position to block any conference in Germany in which Varoufakis tried to participate.

That is one of many reasons why Zoom is an injustice and we should denounce and reject it.

I haven't spoken to the guy since he was a mere hardscrabble multi-millionaire, but this story 100% tracks.

Rick Perlstein:

My first impression of them came of their response to my small-talk description of my delightful afternoon [walking in San Francisco]. Jaws practically dropped, like I had dared an unaccompanied, unarmed stroll through Baghdad's Sadr City in the spring of 2004. [...]

One participant was a British former journalist become computer tycoon who had been awarded a lordship. He proclaimed that the Chinese middle class doesn't care about democracy or civil liberties. I was treated as a sentimental naïf for questioning his blanket confidence.

Another attendee seemed to see politics as a collection of engineering problems. He kept setting up strange thought experiments, which I did not understand. I recall thinking it was like talking to a creature visiting from another solar system that did not have humans in it. [...]

I knew from the New Yorker that Andreessen had grown up in an impoverished agricultural small town in Wisconsin, and despised it. But I certainly was not prepared for his vituperation on the subject. He made it clear that people who chose not to leave such places deserved whatever impoverishment, cultural and political neglect, and alienation they suffered. [...]

And that's when the man in the castle with the seven fireplaces said it.

"I'm glad there's OxyContin and video games to keep those people quiet."

I'm taking the liberty of putting it in quotation marks, though I can't be sure those were his exact words. Marc, if you're reading, feel free to get in touch and refresh my memory. Maybe he said "quiescent," or "docile," or maybe "powerless." Something, certainly, along those lines.

He was joking, sort of; but he was serious -- definitely. "Kidding on the square," jokes like those are called. All that talk about human potential and morality, and this man afire to reorder life as we know it jokingly welcomes chemical enslavement of those he grew up with, for the sin of not being as clever and ambitious as he.

There is something very, very wrong with us, that our society affords so much power to people like this.

Previously, previously, previously, previously.

Modi is building a cult of personality to manipulate the Indian public.

I urge Indians to think deeply and do what Dr Ambedkar would have done.

*Salman Rushdie warns young people against forgetting value of free speech.*

How dangerous, how foolish, and what a shame: young people motivated by good causes think they can make the world better by tactics of bullying. It makes them kin to fascists, and once they eliminate freedom of speech, it is the fascists that take advantage.

The US and Europe have put sanctions on two right-wing Israelis who are leaders of the movement to take Palestinians' land, and associated with pogroms against them.

I don't know how much good this will do, but at least it is an effort in the right direction.

The California thugs who killed Mario Gonzalez by holding him down for a long time face charges of manslaughter.

Car fobs have a security problem. If you’ve parked your car in front of your house someone can relay messages between your key fob and your car, get your car to unlock, get in, and drive off.

This attack is possible because of a sensor problem: The fob and car rely on the strength of the signal between them to sense how far away they are from each other, and that strength can be boosted by an attacker. Thankfully there’s an improved method of sensing distance which is long overdue for being the standard technique, which is to rely on the speed of light. If the fob and car are close enough the round trip time between the two will be low, and if they’re too far away then an intermediary echoing messages can’t reduce the round trip time, only increase it. Thank you absolute speed of light.

As compelling as this is in principle implementing it can be tricky because the processing on the end point needs to be faster than the round trip time. Light goes about a foot in a nanosecond, so you want your total processing time down to a few nanoseconds at the most. This is plenty of time for hardware to do something, but between dodgy and impossible to do any significant amount of computation. But there’s a silly trick for fixing the problem.

Any protocol between the car and fob will end with one final message sent by the fob. To make it round trip secure the fab instead signals to the car that it’s ready to send the final message at which point the car generates a random one time pad and sends it back to the fob, at which point the fob xors the final message with the pad and sends that as the final message. The car can then xor again to get the real final message, authenticate it however is required of the underlying protocol, and if the round trip time on the final message was low enough open up. This allows the fob to calculate its final message at its leisure then load it into something at the hardware level which does xor-and-pong. A similar hardware level thing on the car side can be told to generate a ping with one time pad, then return back the pong message along with a round trip time to receive it. That way all the cryptography can be done at your leisure in a regular programming environment and the low latency stuff is handled by hardware. If you want to be fancy when making the hardware you can even support an identifying code which needs to match on the sending and receiving sides so messages don’t interfere with each other.

Distance detection used on point of sale devices should also work this way. That would have the benefit you wouldn’t have to smush the paying device’s face right into the point of sale machine just to get a reading. The protocol should be a little different for that because in a real payment protocol the paying device should authenticate the point of sale machine rather than the other way around. But the credit card approach does things backwards, and it seems likely that if hardware capability of this sort of thing is built into phones it will be the wrong side.

Someone posted this in a comment on my "sleepy seaside town" post, but this rant goes so hard that I think it deserves its own post:

In my last post (which this post is a superior rehashing of after thinking things over more) I talked about ‘chaos’ which seemed to leave some people confused as to what that meant. Despite being a buzzword which is thrown around in pop science a lot chaos is a real mathematical term with a very pedestrian definition, which is sensitive dependence on initial conditions. It’s a simultaneously banal and profound observation to point out that neural networks as we know them today are critically dependent on not having sensitive dependence on initial conditions in order for back propagation to work properly.

It makes sense to refer to these as ‘sublinear’ functions, a subset of all nonlinear functions. It feels like the details of how sublinear functions are trained don’t really matter all that much. More data, training, and bigger models will get you better results but still suffer from some inherent limitations. To get out of their known weaknesses you have to somehow include superlinear functions, and apply a number of them stacked deep to get the potential for chaotic behavior. LLMs happen to need to throw in a superlinear function because picking out a word among possibilities is inherently superlinear. To maximize an LLMs performance (or at least its superlinearity) you should make it output a buffer of as many relevant words as possible in between the question and where it gives an answer, to give it a chance to ‘think out loud’. Instead of asking it to simply give an answer, ask it to give several different answers, then make give arguments for and against each of those, then give rebuttals to those arguments, then write several new answers taking all of that into account, repeat the exercise of arguments for and against with rebuttals, and finally pick out which if its answers is best. This is very much in line with the already known practical ways of getting better answers out of LLMs and likely to work well. It also seems like a very human process which raises the question of whether the human brain also consists of a lot of sublinear regions with superlinear controllers. We have no idea.

What got me digging into the workings of LLMs was that I got wind that they use dot products in a place and wondered whether the spatial clustering I’ve been working on could be applied. It turns out it can’t, because it requires gradient descent, and gradient descent on top of being expensive is also extremely chaotic. But there is a very simple thing which is sublinear which can be tried: Apply RELU/GRELU to the key and query vectors (or maybe just one of them, a few experiments can be done) before taking their dot product. You might call this the ‘pay attention to the man behind the curtain’ heuristic, because it works with the intuition that there can be reasons why you should pay special attention to something but not many reasons why you shouldn’t.

For image generation the main thing you need is some kind of superlinear function applied before iterations of using a neural network to make the image better. With RGB values expressed as being between 0 and 1 it appears that the best function is to square everything. The reasoning here is that you want the second derivative to be as much as possible everywhere and evenly spread out while keeping the function monotonic and within the defined range. The math on that yields two quadratics, x^2 and its cousin -x^2+2x. There are a few reasons why this logical conclusion sounds insane. First there are two functions for no apparent reason. Maybe it makes sense to alternate between them? Less silly is that it’s a weird bit of magic pixie dust, but then adding random noise is also magic pixie dust but seems completely legit. It also does something cognitively significant, but then it’s common for humans to make a faint version of an image and draw over it and this seems very reminiscent of that. The point of making the image faint is to be information losing, and simply multiplying values isn’t information losing within the class of sublinear functions while square is because if you do it enough times everything gets rounded down to zero.

Frustratingly image classification isn’t iterated and so doesn’t have an obvious place to throw in superlinear functions. Maybe it could be based off having a witness to a particular classification and have that be iteratively improved. Intuitively a witness traces over the part of the image which justifies the classification, sort of like circling the picture of Waldo. But image classification doesn’t use witnesses and it isn’t obvious how to make them do that so a new idea is needed.

DeSantis signs bill, says Satanists can't be school chaplains:

"Some have said that if you do a school chaplain program, that somehow you're going to have Satanists running around in all our schools," he said at a press conference [...]

"Despite DeSantis's contempt for religious liberty, the Constitution guarantees our equal treatment under the law, and DeSantis is not at liberty to amend the Constitution by fiat, at whim," said Lucien Greaves, co-founder of the The Satanic Temple. "He just invited Satanic chaplains into public schools, whether he likes it or not." [...]

"There's some students who need some soulcraft, and that can make all the difference in the world," DeSantis said.

Previously, previously, previously, previously, previously, previously, previously.

Even if there are no legal consequences from any of these trials, it's important to remember that the whole thing just makes him really angry.

And that is a gift that we must treasure.

Trump Rages About His Trial Sketch Artist, Courtroom Nap Reports:

He has privately raged over everything from reports that he can't stop dozing off, to how the court sketch artist is rendering him, to late-night talk show hosts joking about his legal troubles. The former president is reaching levels of fury over the judicial process and all it entails that are "maxed out, even for him," says one source who has had to personally endure Trump's recent rantings about his trial. [...]

Trump has also privately asked people close to him if they agree that the courtroom sketch-artist must be out to get him, two of the sources say. [...]

"I was actually just thinking this morning about how cold [Trump] kept complaining to be yesterday. In his normal world, someone would have jumped up and run, not walked, to get the temperature perfect for him," says Stephanie Grisham, Trump's former White House press secretary who fell out with him years ago. "This entire experience must be beyond uncomfortable for him, not just the fact that it dives into such personal details, but he has absolutely no control for probably the first time since he was a young child."

Grisham adds that Trump being on trial this week "reminded me that not only in the White House, but every facet of his life -- be it at Mar-a-Lago or Trump Tower or Bedminster -- he has a group of people that cater to his every whim. At Mar-a-Lago, people literally stand and applaud him just for walking into the room, and in court, he has to sit there quietly while some people talk about how much they don't like him. I can't imagine how hard it has been for him not to get up and storm out of the place like a five-year-old."

Previously, previously, previously, previously, previously, previously, previously, previously.

A recent scientific expedition to the Gulf of Mexico seafloor shows just how little things have improved near the broken well.

The absence of life is noticeable, says McClain, and what is there doesn't seem healthy. Unlike other wrecks, which tend to become habitats for marine species over time, the sunken Deepwater Horizon has remained comparatively sterile. Organisms that typically inhabit the Gulf's seafloor -- such as sea cucumbers, giant isopods, corals, and sea anemones -- are simply missing, says McClain. Perhaps more concerning are the crabs. Naturally red, the crabs McClain and his team pulled up in their traps were tinted an oily black; many were also missing legs, while others had lesions. [...]

Benfield had visited the oil spill site before, including shortly after the explosion in 2010. He also joined the first scientific research expedition to the exploded wellhead in 2017. When Benfield saw the site then, he was shocked by how little it had recovered. Another seven years on, Benfield says researchers are slowly starting to see more animals. There's "more diversity of fishes and macroinvertebrates," he says. But compared with before the explosion, the site remains a desert, Benfield adds.

For those who embarked on the most recent expedition, the dire sight has them questioning how the Gulf will fare in the future. "We may not actually ever see recovery," [...] "Maybe in my kid's lifetime. But it's going to take a long time, I think."

Previously, previously, previously, previously, previously, previously, previously.

"Post dot news", the Andreessen-funded cryptocurrency grift masquerading as a social network, that considered dunking on billionaires to be hate speech, and that created fake "placeholder" accounts to try and get their users to bully news organizations into signing up... is shutting down.

Something something "incredible journey".

Previously, previously, previously, previously, previously, previously, previously, previously, previously.